Only allow SOMATIC or GERMLINE values for SV_Status column

Author: oplantalech

scripts/importer/validateData.py

@@ -3217,8 +3217,9 @@ def checkNCBIbuild(ncbi_build):
         def checkSVstatus(sv_status):
             """ Allowed values for SV_Status column are SOMATIC and GERMLINE """
             if sv_status is None:
-                self.logger.warning('No value in SV_Status, assuming the variant is SOMATIC',
-                                  extra={'line_number': self.line_number})
+                self.logger.error('Invalid SV_Status value: possible values are [SOMATIC, GERMLINE]',
+                                  extra={'line_number': self.line_number,
+                                         'cause': sv_status})
             elif sv_status.lower() == 'germline':
                 self.logger.warning('GERMLINE variant identified from the SV_Status value. If this variant is not meant for public release, please remove it.',
                                   extra={'line_number': self.line_number,

tests/unit_tests_validate_data.py

@@ -2023,9 +2023,15 @@ def test_missing_values(self):
         self.logger.setLevel(logging.ERROR)
         record_list = self.validate('data_structural_variants_missing_values.txt',
                                     validateData.StructuralVariantValidator)
-        self.assertEqual(2, len(record_list))
+        self.assertEqual(3, len(record_list))
         record_iterator = iter(record_list)
 
+        # Expected first error due to a SV_Status being empty
+        record = next(record_iterator)
+        self.assertEqual(logging.ERROR, record.levelno)
+        self.assertEqual(3, record.line_number)
+        self.assertIn("Invalid SV_Status value: possible values are [SOMATIC, GERMLINE]", record.message)
+
         # Expected ERROR message due to missing Entrez gene id and/or gene symbol at site 1 and site 2
         record = next(record_iterator)
         self.assertEqual(logging.ERROR, record.levelno)